CT radiation risks coming into clearer focus

BMJ 2013; 346 doi: http://dx.doi.org/10.1136/bmj.f3102 (Published 21 May 2013) Cite this as: BMJ 2013;346:f3102
  1. Aaron Sodickson, section chief of emergency radiology and medical director of computed tomography
  1. 1Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA
  1. asodickson{at}partners.org

As we gather more direct evidence of the dose-response curve from childhood CT

Recent attention to the cancer risks of ionizing radiation has prompted vigorous debate about how to quantify the risks of diagnostic imaging, and whether or how these risks ought to be incorporated into our decision making process as we participate in patient care.

In the past, models of the carcinogenic risks of ionizing radiation have primarily relied on long term surveillance of the Japanese atomic bomb survivors, which showed significant increases in the incidence of cancer after effective doses greater than about 50 mSv.1 2 The relative paucity of direct data in the lower dose range delivered by diagnostic imaging has led to conflicting opinions about the shape and slope of the radiation dose-response curve.

In a linked study (doi:10.1136/bmj.f2360), Mathews and colleagues present compelling data on the magnitude of the cancer risk attributable to ionizing radiation.3 This well designed study examined a cohort of nearly 11 million young patients in the Australian national Medicare system and compared subsequent incidence of cancer in the 680 000 patients exposed to computed tomography (CT) with that in unexposed controls.

The finding that will probably dominate media headlines is that exposure to CT in childhood increased the incidence of cancer by 24%. However, it is important to recognize that the baseline incidence of cancer in a general pediatric population is extremely small, so that a 24% increase makes this risk just slightly less small. To put these numbers in context, it is necessary to consider absolute (rather than relative) cancer risk, and to relate the increase to the degree of exposure. The authors found an overall excess risk of about 0.125 cancers per Sievert, which equated to roughly one excess cancer per 1800 head CTs (each with an average estimated dose of around 4.5 mSv). This would equate to roughly one excess cancer per 4000 head CTs at the more typical doses in use with current day technology (around 2 mSv).

Mathews and colleagues compared their results with those of the Life Span Study of Japanese atomic bomb survivors1 and those of the more recent landmark UK study.4 All three studies show good concordance within the confidence intervals that their cohort sizes permit. Although the UK study was powered to detect significant increases in childhood leukemia and brain tumors,4 the current study, which is larger, shows significantly increased risks across a large range of cancer types.

This observed increase in risk associated with the low radiation doses delivered by CT scans supports the most widely adopted linear-no-threshold dose-response model in which double the radiation dose is assumed to impart double the cancer risk. The reported risks also roughly match the lifetime attributable risks predicted by the BEIR-VII (biological effects of ionizing radiation) report,5 one of the most commonly used linear-no-threshold models.

So what should physicians do with this information and how can it be incorporated into practice? There are many possible interventions to control patients’ exposure to radiation, which can conceptually be grouped into timeframes—before, during, and after the CT scan.

Before the scan, there are many opportunities to control the use of imaging. Although the clinical benefits of a medically indicated scan usually far outweigh the small associated risk of developing cancer, this is the time for critical assessment of what impact the imaging result might have on the patient’s care plan. Special attention should be paid to patients undergoing recurrent imaging, because if frequently repeated scans are found to provide little clinical benefit, the cumulative risk-benefit balance may support a decision not to image again for the same clinical presentation. Imaging algorithms or evidence based clinical decision rules may be adopted for clearly defined clinical scenarios. Electronic decision support embedded in the scan ordering process can substantially reduce utilization.6

During the scan, there are many available methods to reduce the radiation dose without negatively affecting the diagnostic quality of the examination.7 Although CT radiation doses vary considerably, existing dose reduction tools and ongoing technological improvements allow CT scans to be performed using substantially lower radiation doses than was possible with previous generations of scanners, such as those in place during the period of the current study.8 Improved adoption of such tools is key, through collaborative efforts of radiologists, CT manufacturers, medical physicists, and CT technologists.

After the scan, opportunities for managing radiation dose are increasing. Adoption of newly developed informatics methods that enable large scale data capture of CT scanner radiation output has resulted in databases that will be vital for institutional benchmarking, optimization of CT protocols, and quality control.9 Such data capture would also enable more accurate patient specific dose estimation than was possible from the data sources available to Mathews and colleagues.3 Although the authors assigned credible doses to the scans in their study, future epidemiological work may be greatly enhanced by improved capture of patient specific dosimetry.

With further validation of radiation risk models, not only in children but also in adults, we will ultimately be able to perform more accurate patient specific risk assessment to better inform imaging decisions. Mathews and colleagues’ study is a vital step towards this goal.3


Cite this as: BMJ 2013;346:f3102


  • Research, doi:10.1136/bmj.f2360
  • Competing interests: I have read and understood the BMJ Group policy on declaration of interests and declare the following interests: I have served as a consultant for Siemens CT and Medrad (now Bayer) within the past three years.

  • Provenance and peer review: Commissioned; not externally peer reviewed.